The present study was a retrospective, single center, observational study of consecutive AMI patients who received VA-ECMO for refractory CS or cardiac arrest between May 2004 and April 2016 at Samsung Medical Center. Patients 18 years or older were included in this analysis. Venoarterial-ECMO was indicated in AMI patients with cardiac arrest or acute refractory cardiovascular failure, defined as evidence of organ hypoperfusion (extensive skin mottling, progressive lactic acidosis, oliguria or altered mental status) despite adequate intravascular volume management and maximal medical therapy including vasopressors or inotropes. All patients were expected to undergo early revascularization and to receive optimal medical therapy in accordance with current guidelines.15,16 A total of 145 patients with AMI who received VA-ECMO were finally eligible for this study. The final decision to implant VA-ECMO was determined by an experienced team, and the VA-ECMO was inserted at the bedside or in a catheterization laboratory by cardiovascular surgeons or interventional cardiologists. The requirement for informed consent was waived by the Institutional Review Board of Samsung Medical Center because of the retrospective nature of the study.

Details of the management of VA-ECMO for AMI patients have been previously documented.18 In brief, a VA-ECMO device was inserted by percutaneous cannulation using the Seldinger technique or surgical cannulation using the cut-down method. Arterial cannula sized 14-Fr to 17-Fr and venous cannula sized 20-Fr to 24-Fr were used. Femoral vessels were usually used as vascular access sites. Capiox Emergency Bypass System (Capiox EBS; Terumo, Inc, Tokyo, Japan) and Permanent Life Support (PLS; MAQUET, Rastatt, Germany) were available in our hospital at the time of the study. In the event of distal limb ischemia after arterial cannulation, a catheter was inserted distal to the cannulation site for limb perfusion. Pump speed was adjusted to obtain a cardiac index greater than 2.2 L/min/body surface area (m2), mean arterial pressure> 65mmHg, and central mixed venous saturation> 70%. Intravenous heparin was infused to maintain an activated clotting time ranging from 150 to 180seconds unless life-threatening bleeding was observed. Successful weaning was defined as disconnection of the patient from ECMO without reinsertion or death within 24hours.

The following information was collected retrospectively through medical record review: age, sex, physical examination results, underlying comorbidities, angiographic data, laboratory data, echocardiography data, in-hospital management, and pre- and post-ECMO data. When the same laboratory data were measured several times before ECMO insertion, the laboratory value measured at the nearest time to ECMO insertion was recorded. When the neurologic evaluation was available before ECMO insertion, the Glasgow coma score was obtained at the nearest time to ECMO insertion time. However, if the patient was unable to perform a neurologic evaluation due to unexplained or out-of-hospital cardiac arrest, the Glasgow coma score was calculated after ECMO insertion. Scores from SAVE, and ENCOURAGE, which are previously validated scoring systems for evaluating survival to discharge in patients undergoing VA-ECMO, were calculated for comparison with the newly generated risk prediction model.17,19 Additional clinical information including follow-up data were obtained from medical records and telephone interviews by trained study coordinators if necessary. The median follow-up duration was 33.0 days [interquartile range from 4.0 to 373.0 days]. The primary outcome was in-hospital mortality, while the secondary outcome was all-cause mortality during the follow-up period.

Categorical variables are presented as numbers and relative frequencies and their group differences were compared using the chi-square test or the Fisher exact test, as appropriate. Continuous variables are presented as mean±standard deviations or medians (25th-75th percentiles) and their group differences were compared using the Student t test. Multivariable logistic regression analysis was used to identify predictors of in-hospital mortality in patients with AMI who underwent VA-ECMO implantation. For practical utility purposes, continuous variables were transformed into categorical variables, which were assessed by normal range or best cutoff values in the maximally selected minimum P value method. Covariates that were considered clinically relevant or that showed a univariate relationship with outcome (P ≤ .2) were entered into the multivariable logistic regression model. Clinically intercorrelated variables were excluded from the model. Then, stepwise elimination analysis was performed to identify a useful subset of predictors. The AMI-ECMO risk score was then constructed to predict in-hospital mortality using a regression coefficient-based scoring method. To generate a simple integer-based point score for each predictive variable, each β coefficient was divided by the absolute value of the smallest coefficient, multiplied by 5, and rounded to the nearest integer. The discriminating power of the constructed score in predicting the in-hospital mortality was assessed by considering the area under the curve from the receiver operating characteristic (ROC) analysis. The adequacy of the model was checked using the Hosmer-Lemeshow goodness-of-fit test, and leave-one-out cross-validation was used for internal validation to obtain the misclassification error rate. To evaluate the association between AMI-ECMO score and estimated risk of all-cause mortality, the probability of death was estimated using the Cox proportional hazards model. The estimated risk of all-cause mortality was depicted using a locally weighted scatterplot smoothing regression line. The property of AMI-ECMO score was then compared with those of the ENCOURAGE and SAVE scores. Event-free survival according to the AMI-ECMO score quartiles was evaluated by Kaplan-Meier analysis, and the significance level was assessed with a log-rank test. Correlations among AMI-ECMO, ENCOURAGE, and SAVE scores were assessed with the Spearman correlation coefficient. Statistical analyses were performed using R Statistical Software (version 3.2.5; R Foundation for Statistical Computing, Vienna, Austria) with P <.05 considered statistically significant.

Among 145 AMI patients treated with VA-ECMO, 97 (66.9%) patients presented with ST-segment elevation myocardial infarction. Successful ECMO weaning was achieved in 91 patients (62.8%), and 76 patients (52.4%) survived to discharge. The median duration of mechanical support was 2.0 days [interquartile range: 1.0-4.0]. Table 1 shows the baseline clinical and laboratory characteristics of the participants, and Table 2 shows angiographic findings and provides in-hospital management information. Compared with survivors, nonsurvivors were older and had a lower systolic blood pressure, pulse pressure, heart rate, Glasgow coma score, proportion of successful revascularization, and lower return of spontaneous circulation before ECMO insertion. The incidence of gastrointestinal bleeding and lactic acid values were significantly higher in nonsurvivors than in survivors. Among the study population, only 1 patient received heart transplantation during VA-ECMO maintenance. However, heart transplantation was performed in 6 patients during follow-up because of aggravation of ischemic cardiomyopathy after AMI.

After continuous variables were transformed into categorical variables, independent predictors of in-hospital mortality were age> 65 years, body mass index> 25 kg/m2, Glasgow coma score <6, lactic acid> 8 mmol/L, anterior wall infarction, and no attempted or failed revascularization, based on multivariable analysis with adjustment for prognostic covariates (Table 3). The AMI-ECMO score was then generated to predict in-hospital mortality in AMI patients who received VA-ECMO. A full description of the model is provided in Table 4. The incidence of in-hospital mortality after VA-ECMO device implantation was 6.2%, 28.1%, 51.6%, and 93.8% for AMI-ECMO scores 0 to 16, 17 to 19, 20 to 26, and> 26 (divided by quartiles), respectively (P <.001 for trend). In addition, the AMI-ECMO score had good predictive ability as assessed by the area under the curve in ROC curve analysis (C-statistic 0.880; 95% confidence interval [95%CI], 0.820-0.940) with satisfactory calibration (Hosmer-Lemeshow chi-square=7.484, df=8; P=.485) (Figure 1). Compared with previously validated scoring systems, such as ENCOURAGE or SAVE, the AMI-ECMO score had the highest predictive ability in the present study population (C-statistic for AMI-ECMO score vs ENCOURAGE score, 0.880 vs 0.756; P=.003, ROC and AMI-ECMO score vs SAVE score 0.880 vs 0.647; P <.001, respectively). Correlations among scoring systems are shown in Figure of the supplementary material and demographic differences between AMI-ECMO and ENCOURAGE are shown in the Table of the supplementary material. Leave-one-out cross-validation (analysis of the AMI-ECMO score for internal validation showed that it had the lowest misclassification error rate of in-hospital mortality of 14% compared with 26% and 38% for ENCOURAGE and SAVE, respectively.

Figure 1.

Receiver operating characteristic curves of AMI-ECMO, ENCOURAGE, and SAVE scores for predicting in-hospital mortality. Receiver operating characteristic curves for predicting in-hospital mortality based on AMI-ECMO, ENCOURAGE, and SAVE scores are presented. Among the models, the AMI-ECMO score showed the highest C-statistic value. AMI, acute myocardial infarction; ECMO, extracorporeal membrane oxygenation.

(0.1MB).

Among the patients who survived to discharge (76 patients), 8 patients died, 1 patient was lost to follow-up, and 68 patients were still alive during the follow-up period. AMI -ECMO scores as continuous values showed a significant association with the estimated rate of all-cause mortality (per 1 increase, hazard ratio, 1.11; 95%CI, 1.08-1.14, P <.001) (Figure 2). Comparison of 60-day all-cause mortality rates among the 4 groups, classified according to AMI-ECMO score quartiles, are shown in Figure 3. There were significant differences in all-cause mortality among the 4 groups (log-rank P <.001).

Figure 2.

Estimated all-cause mortality rates according to AMI-ECMO score. The locally weighted scatterplot smoothing curve depicts correlations between the probability of all-cause mortality, which was estimated using a Cox proportional hazards regression model, and the AMI-ECMO score. The AMI-ECMO score was significantly correlated with estimated rate of all-cause mortality. 95%CI, 95% confidence interval; AMI, acute myocardial infarction; ECMO, extracorporeal membrane oxygenation; HR, hazard ratio.

(0.09MB).

Figure 3.

Comparison of all-cause mortality according to AMI-ECMO score quartiles. Kaplan-Meier curve of 60-day survival probability according to AMI-ECMO score quartiles (≤ 16, 17 to 19, 20 to 26,> 26) is plotted. AMI, acute myocardial infarction; ECMO, extracorporeal membrane oxygenation.

(0.08MB).

Our study has several limitations. First, due to the retrospective nature of our database, some variables including lactic acid level, echocardiographic parameters, hemodynamic parameters, and neurologic data were not recorded for all patients. In particular, serum lactate data were available for only 87.6% (127/145) of the patients. Therefore, cases with missing value of serum lactate were excluded from the multivariable logistic regression model. Second, the AMI-ECMO score needs to be prospectively confirmed in other patient populations with AMI complicated by CS treated with VA-ECMO, because we performed only an internal validation of the AMI-ECMO score. Finally, this study was based on a single-center experience, which may limit the generalizability of our results and have created selection bias in the patients’ subset. Also, the relatively small sample size could have limited the precision of the estimates.